Occupational hearing loss remains a problem, despite the efforts made by implementing hearing conservation programs in the workplace. The first issue is that the actual passive noise reduction of the hearing protector worn during the work shift greatly differs from the optimal passive noise reduction measured in the laboratory due to suboptimal placement, inconsistent use and in general variations in the acoustical seal over time. Despite the recent development of a field attenuation measurement system for hearing protection devices, the precise residual noise level under the hearing protector remains unknown. The second issue is that, even if this individual noise exposure would be known precisely for each worker, the effective risk of hearing damage would still remain uncertain given the difference between worker's susceptibility to develop noise-induced hearing loss.
To address simultaneously these two issues, an alternative approach would consist in measuring the auditory health changes induced by daily noise exposure on an individual basis and to immediately warn the worker (in real-time) when a change in hearing sensitivity is taking place, before any permanent damage is caused. In clinical practice, a wide range of audiological tests are available to assess hearing status. However, with respect to occupational noise exposure, these tests are not conducted frequently enough for early detection of changes in hearing sensitivity induced by noise exposure, and also not sufficiently robust to be carried out in an environment where acoustical and electrical noise intensity levels are too high. Moreover, the whole procedure to monitor a worker's hearing health daily takes too much time for most standard audiological tests and would interfere with the worker's work routine.
Indeed, distortion product otoacoustic emissions (DPOAEs) offer an objective, fast and reliable way to detect early signs of noise-induced changes in hearing sensitivity. When two pure tone stimuli, f1 and f2 with the f2/f1 ratio typically around 1.22, are sent through the two miniature receivers of the otoacoustic emission (OAE) probe, low-level cubic distortion signals (i.e. fdp=2f1−f2) are generated by an active non-linear process inside the inner ear. These signals travel back from the inner ear to the outer ear canal where they can be recorded. If the outer hair cells inside the cochlea of the inner ear are damaged—for instance due to previous excessive noise exposure—the amplitude of DPOAEs is found to be lower than if they would be healthy.
Nevertheless, various clinical test setups for DPOAEs have been commercially available for more than 15 years, now ranging from standalone all-in-one hand-held devices to more advanced systems with two probe measurement interfaces connected to a personal computer. No commercial system currently on the market can continuously monitor DPOAEs in a given individual, in field conditions, because the nominal DPOAE signal, generally at levels between −5 dB to 20 dB sound pressure levels (SPL) is disturbed by the background noise. Proper recording of DPOAE responses is very vulnerable to interfering background noise which normally largely exceeds these low level responses.
In the case of hardware solutions, standard probe eartips usually provide a certain amount of passive noise reduction, but this noise reduction is not individually optimized and it is not sufficient for noisy test environments. Even though passive earmuffs could be used on top of the DPOAE probe, they may not provide sufficient additional low frequency attenuation in order to measure DPOAEs accurately in industrial environments. Unfortunately, placing an earmuff on top of an OAE probe might slightly dislocate the probe and hence require more strict supervision of calibration procedures. This situation would conflict with the final aim of OAE monitoring without any external supervision.
In the case of software solutions, the standard noise rejection techniques and time averaging can improve the signal-to-noise ratio (SNR) in case of limited disturbance, but this has shown to be insufficient in more realistic occupational noise settings. Moreover, these techniques do not offer sufficient improvement to lower the noise floor in the frequency range below 1500 Hz to measure DPOAEs accurately even in lower background noise levels.
In response to the problems encountered with averaging methods, several adaptive filtering techniques have been studied. Delgado's adaptive filtering technique uses a contralateral internal ear microphone (IEM) as a physiological noise reference and an ipsilateral outer ear microphone (OEM) as an external background noise reference to remove the noise captured in the tested ear IEM. This adaptive filtering algorithm was proven to increase the SNR on the whole frequency spectrum while reducing the test time needed by normal time averaging methods. Although Delgado's adaptive filtering does lower the noise level in the DPOAE signal, it has not been tested in realistic noise conditions and a somewhat low signal-to-noise ratio improvement was obtained with laboratory setup experiments.
Furthermore, although additional passive noise reduction and other hardware improvements might improve the Signal-to-Noise ratio (SNR), studies have shown that in order to extract the level of the DPOAE signal in a noisy environment, a more robust signal processing scheme is needed.
In order to extract the level of the DPOAE signal after the noise rejection processing, a robust sinusoid (or tonal) signal extraction algorithm is needed. Previous studies have shown a promising approach to extract DPOAE signals without the need of a Fast Fourier Transform (FFT). This approach is more robust to higher noise levels and, since it is not FFT based, it can be used at any stimuli frequency (respecting the 1.22 ratio) without having to keep an integer multiple of the frequency resolution (Δf) of the FFT to reduce the spectral leakage. The extended stimuli frequency range capabilities of such an algorithm, may give the opportunity for researchers to characterize the cochlea with a finer frequency resolution than FFT based algorithms.
Also, it has been shown that a previously proposed algorithm is very sensitive to the adjustment of various parameters i.e. filter adjustments, adaptation step sizes and normalization gains in order to optimize the algorithm for all possible situations such as, for example, different DPOAE magnitudes and various noise conditions. Therefore, such an algorithm is not practical to assess a worker's cochlea functionality in an automatic and autonomous manner if parameters need to be changed constantly.
While the noise rejection algorithm and the involved hardware does lower the noise floor and increase the DPOAE level reliability, the alternative to the FFT based DPOAE level extraction can also contribute to reduce the noise level since the FFT is very sensitive to background noise in frequency bins near the DPOAE frequency and the stimuli. The magnitude of the stimuli, noise signals and DPOAE response causes spectral leakage around the DPOAE frequency which introduces an error in the estimation of the DPOAE level when the DPOAE frequency is not an integer multiple of the frequency resolution (Δt). A method to solve these problems has been proposed to extract non-stationary sinusoids with a non FFT based algorithm, but is too sensitive to the adjustment of various parameters and therefore unrealistic to use.
According to the results obtained in previous studies, there is a need for an improved method and device for continuous in-ear hearing health monitoring on a human being based on measurements of otoacoustic emissions (OAE) in order to use such device for in-field applications.